So it's not directly audio related, but I figured someone around here would have a good idea where to start.
I need to amplify a sine wave (maximum of 1 V, around 20 Mhz) to 100 V amplitude (200 Vpk-pk) to drive a capacitor of about 30 pF (not really a capacitor, but a variable diffraction crystal, hooked up like a capacitor).
I know the design should be pretty simple (aside from the 200 V supply) since the power is so low (almost 0?), but it seems like no RF rated components go near 200v rails… I found an op-amp with a 300v supply rating and 175 Mhz GBP, but the slew rate was only 200 V/uS (needs to be > 8000 V/uS). I could use a lower power op-amp and a transformer, but then we are locked on 20 Mhz and there’s all sorts of goofy LRC stuff to worry about with the capacitor. So I started looking at building a class AB amplifier, but at RF it takes 4 or 5 or 6 stages to get close to the 200x gain required, and the distortion would be horrendous (0.1%THD is nowhere near good enough for 20 Mhz)…
Can I just use a 300 V MOSFET with a large load resistor (200k) in parallel with the (load) capacitor? Or is that making things too easy?
I need to amplify a sine wave (maximum of 1 V, around 20 Mhz) to 100 V amplitude (200 Vpk-pk) to drive a capacitor of about 30 pF (not really a capacitor, but a variable diffraction crystal, hooked up like a capacitor).
I know the design should be pretty simple (aside from the 200 V supply) since the power is so low (almost 0?), but it seems like no RF rated components go near 200v rails… I found an op-amp with a 300v supply rating and 175 Mhz GBP, but the slew rate was only 200 V/uS (needs to be > 8000 V/uS). I could use a lower power op-amp and a transformer, but then we are locked on 20 Mhz and there’s all sorts of goofy LRC stuff to worry about with the capacitor. So I started looking at building a class AB amplifier, but at RF it takes 4 or 5 or 6 stages to get close to the 200x gain required, and the distortion would be horrendous (0.1%THD is nowhere near good enough for 20 Mhz)…
Can I just use a 300 V MOSFET with a large load resistor (200k) in parallel with the (load) capacitor? Or is that making things too easy?
bandwidth and distortion requirements affect the choice but tuned transformer circuits would get the 200 V from more cheaply available lower V SS amps
applying negative feedback for distortion reduction requires considerably higher frequency response than the working frequency and TO-220/247 series parts have parasitic lead inductance that limit frequency response
IXYS RF has power fets that perhaps could be the basis of a direct drive output amplifier
IXYS RF: HF/VHF Linear MOSFETs
applying negative feedback for distortion reduction requires considerably higher frequency response than the working frequency and TO-220/247 series parts have parasitic lead inductance that limit frequency response
IXYS RF has power fets that perhaps could be the basis of a direct drive output amplifier
IXYS RF: HF/VHF Linear MOSFETs
Does this need to frequency modulate, or is it fixed? Either way resonance may be your friend...learn about matching and Smith charts. I used to design Class-E generators that would do 250W into 50 Ohm loads, off of a 24V supply rail.
-SpeakerScott
P.S. This is a non-trivial effort....take your time. If you do not have access to network analyzer then it will be made more difficult.
-SpeakerScott
P.S. This is a non-trivial effort....take your time. If you do not have access to network analyzer then it will be made more difficult.
Last edited:
how about having the 30pf crystal as capacitive element in a series resonant LC circuit, driven by a low voltage high current amp?
this amp can then even be quite non linear, and the resonant circuit will smooth it out.
i could do this i think ;p
(edit: i am not alone i see lol 🙂
and, that is some slew rate ;p i quess that implies broadband functionality (as in not tuned or resonant)
this amp can then even be quite non linear, and the resonant circuit will smooth it out.
i could do this i think ;p
(edit: i am not alone i see lol 🙂
and, that is some slew rate ;p i quess that implies broadband functionality (as in not tuned or resonant)
Last edited:
You could probably modify a CRT video projector CRT driver card.
This one does a least a 80MHZ square at 200V P-P and over 100 MHZ sine into 10pf CRT cathode.
Where 1V input will give you 0V output, 0.075V input will give you 200V output and -0.4V will give you 260V output.
This one does a least a 80MHZ square at 200V P-P and over 100 MHZ sine into 10pf CRT cathode.
Where 1V input will give you 0V output, 0.075V input will give you 200V output and -0.4V will give you 260V output.
Last edited:
I'd never apply a network analyzer to a circuit when I've got a 200V supply.
This would be quite a hard work. First af all you need a good starting point for your amplifier, as this makes life much easier. Maybe the CRT amp isn't bad. I'd use a spice simulator and start tweaking that amp.
I think some days ago I found an application note from IXYS Power. Don't konw which one it was exactly.
If connecting to your power circuit, attenuators are a must. If trying to analyze your load, and you have reason to believe that your load is linear with power....then you can use it to directly characterize the load.
BTW...what the heck is this 20MHz for?
-S
A differential E810F (a.k.a. 7788) in triode mode. Should take 20 MHz in it's stride. A single tube will give you a gain of x50, so differential is your answer. If you need low output impedance then 'mu-follower' the tubes with FETs, or add some form of solid state follower output buffer.
: )
: )
Thanks for the input. It seems this is more complicated than I had originally thought.
The company that supplied the crystal offers a resonant version with a 16-22 V input for a specific frequency so it's clearly possible to do it that way. I'd like to stay away from a tank circuit because any change in the frequency requires a re-design. I'm helping the physics department with this (I'm just the EE) so I'm not sure exactly what the crystal does, but it's got to do with feedback from a laser diode to keep it's frequency locked in. The frequency that the crystal runs at is the same as the frequency that the laser runs at, and if they change the laser frequency it would be nice to keep the same amplifier.
As for the first fellow that posted, yes, you're right. By my maths an ideal 30 pF at 20 Mhz is 265.25 ohms reactance which at 100 VAC should drop 266 mA or 18.8 W. A push-pull (AB) amplifier would need two transistors rated to dissipate around 30 W for a good safety factor in cooling.
Such transistors exist, eg ON # MJW1302AG and # MJW3281AG ( http://www.onsemi.com/pub/Collateral/MJW3281A-D.PDF ), or MJL4302AG and MJL4281AG ( http://www.onsemi.com/pub_link/Collateral/MJL4281A-D.PDF ) which would just need a decent driver and a +- 100 V supply. The gain would be relatively low (1.5 - 30 Mhz GBP) but it is much easier to find a high current, low voltage, HF op-amp than a low current, HV, HF op-amp. The problem then comes into building a low distortion AB amplifier with 20+ Mhz bandwidth - and it's hard enough to do that with audio frequencies.
The CRT idea is promising - an old scope or TV would also likely have the necessary supply transformer. I actually have one of these half taken apart in my shop, I'll go through it this afternoon (no need for the safety lecture, I've got it 😉. A tube would be very elegant - but a very large one would be required and I think SS would be more efficient.
In the spirit of KISS, it is tempting to try a pure class A... Dumping 20w+ into a heatsink with a fan ( like http://www.diyaudio.com/forums/diyaudio-com-articles/160464-de-lite-amplifier.html?garpg=4 ).
The company that supplied the crystal offers a resonant version with a 16-22 V input for a specific frequency so it's clearly possible to do it that way. I'd like to stay away from a tank circuit because any change in the frequency requires a re-design. I'm helping the physics department with this (I'm just the EE) so I'm not sure exactly what the crystal does, but it's got to do with feedback from a laser diode to keep it's frequency locked in. The frequency that the crystal runs at is the same as the frequency that the laser runs at, and if they change the laser frequency it would be nice to keep the same amplifier.
As for the first fellow that posted, yes, you're right. By my maths an ideal 30 pF at 20 Mhz is 265.25 ohms reactance which at 100 VAC should drop 266 mA or 18.8 W. A push-pull (AB) amplifier would need two transistors rated to dissipate around 30 W for a good safety factor in cooling.
Such transistors exist, eg ON # MJW1302AG and # MJW3281AG ( http://www.onsemi.com/pub/Collateral/MJW3281A-D.PDF ), or MJL4302AG and MJL4281AG ( http://www.onsemi.com/pub_link/Collateral/MJL4281A-D.PDF ) which would just need a decent driver and a +- 100 V supply. The gain would be relatively low (1.5 - 30 Mhz GBP) but it is much easier to find a high current, low voltage, HF op-amp than a low current, HV, HF op-amp. The problem then comes into building a low distortion AB amplifier with 20+ Mhz bandwidth - and it's hard enough to do that with audio frequencies.
The CRT idea is promising - an old scope or TV would also likely have the necessary supply transformer. I actually have one of these half taken apart in my shop, I'll go through it this afternoon (no need for the safety lecture, I've got it 😉. A tube would be very elegant - but a very large one would be required and I think SS would be more efficient.
In the spirit of KISS, it is tempting to try a pure class A... Dumping 20w+ into a heatsink with a fan ( like http://www.diyaudio.com/forums/diyaudio-com-articles/160464-de-lite-amplifier.html?garpg=4 ).
No, that will not work. First the input capacitance is much too high. Second a ft=30Mhz means you will nearly have no gain at 20MHz. You'll need a transistor with a Ft of some 100s Mhz and also ultra-low Cbc. Maybe something like 2SA1381/2SC3503 and parallel two or three of them. Or a MOSFET with a very low Cgd.
It is a non-trivial effort, but class E is best designed/tuned with an oscilloscope, not a PNA. At 20 MHz, you can see the waveforms and tune it right in. Explicit equations for class E exist, but Cripps' book has a tuning procedure if there are any unknowns (like what *is* RL - if you don't specifically have a measurement of your output tranny Q....)
The other benefits of class E is that it remains very low in harmonic output (distortion) even when it is not fully saturated. 3rd order IM is pretty terrible, but if it's a single tone it's the way to go. If you're amplitude modulating in any way, you want class F - and that has an explicit design procedure, too.
I like the MOSFET (al la original post), but I'm not quite sure of how "low" the gate capacitance needs to be.
Would something like the Infineon SPD01N60C3 with 100 pF input capacitance work? At 11w power dissipation for Class A I'd obviously need 4 or more...
The effective input capacitance must be lower than the load capacitance. As the transconductance for a given current is higher for a bipolar transistor, I prefer this kind of type here. Low capacitance means the lowest you can find. Otherwise you have to design a (broadband) rf amplifier with load matching techniques. 
I think something like this could be a good starting point. Try to simulate this with spice and see if it matches your requirements.
I think something like this could be a good starting point. Try to simulate this with spice and see if it matches your requirements.
The effective input capacitance must be lower than the load capacitance. As the transconductance for a given current is higher for a bipolar transistor, I prefer this kind of type here. Low capacitance means the lowest you can find. Otherwise you have to design a (broadband) rf amplifier with load matching techniques.
I think something like this could be a good starting point. Try to simulate this with spice and see if it matches your requirements.
Ran some SPICE and it's gonna take a lot of stages to get the gain. Played with other transistors too. I guess tank circuit it is... 🙁
Thanks guys,
Jacob
When you set a (high-speed) op in front of the basic circuit, set the right values to the resistors, apply the right operating points and apply some feedback, this stage should be roughly able to solve your problem. 😉
Buy a shortwave communications transceiver in the 50W range with an antenna tuner. Parallel (series?) your load with a resistive dummy load in the 250 ohm region, connect it all with some co-ax, set it to 20MHz CW, key it and let the tuner do it's thing. Something like that.
w
Don't forget to take the radio home when its all over.
w
Don't forget to take the radio home when its all over.
Last edited:
No, not really. Just one stage for the gain (a differetial triode connected 7788), then add a second stage of solid state followers. No point to make it more difficult. Up to you. Good luck. 🙂
Er, no, we haven't used valves for this kind of stuff for 30 years or more. You can't hear 20MHz. Using valves just complicates things, makes it more expensive and dangerous, you need a chassis, valve bases, hand-made tank coil with tap depending on design, HV caps, etc., etc.
Use an off-the-shelf solid-state (1.8-30MHz) linear HF amplifier. You can drive it with a sig. gen. and turn the output up and down 'till you get 200V Pk-Pk measured at the load. You won't need any matching if you parallel your capacitive load with a 50 ohm dummy load, which will pretty much swamp the capacitor and prevent too high an SWR on the cable and the destruction of the output devices. (Maybe).
You can probably save some money (not much) by building a narrowband amp with a tank, but you won't learn how to design it (or the broadband one with TL transformers) in under 6 months. Just identifying and obtaining the right ferrites can be a big problem, and even then you'll only be reworking an existing prototype.
If you leak signal you'll almost certainly be breaking the law.
w
- Status
- Not open for further replies.